New Relic Case Study

bolivialodgeInternet και Εφαρμογές Web

14 Δεκ 2013 (πριν από 3 χρόνια και 7 μήνες)

94 εμφανίσεις

CanvasPop Cuts Application
Response Time to Microseconds
with Help from New Relic
Since 2009, CanvasPop has delivered more than
280,000 high quality, custom-made canvas photo prints
to 100,000+ customers across North America. With an
“obsessed” approach to customer service — backed by
an extra-ordinary lifetime guarantee — the company
matches every customer with a personal in-house
designer to help ensure total satisfaction.
CanvasPop operates two facilities in North America.
One is the company’s headquarters in Ottawa, and the
other is a 20,000 square foot facility in Las Vegas built
to support manufacturing and order fulfillment across
the continent.
Environment
The vast majority of CanvasPop’s code-based infrastructure is hosted on Amazon Web Services. And PHP is the code
of choice for most applications. The primary customer-facing application instance is a single Multi-AZ load balancer
with a two-node cache cluster and a Multi-AZ instance on Amazon RDS.
Other components of CanvasPop’s predominantly open source based environment include:
• Linux
• Apache HTTP Server
• Chef systems integration framework
• Git source control
• Gearman job queue server
• jQuery
• Backbone.js
• Zend Framework
• ImageMagick
At a Glance
INDUSTRY
e-commerce
LOCATION
Ottawa, Ontario, Canada
EMPLOYEES
50
USE CASE
Monitor performance on
customer-facing dot-com
application and internal
manufacturing application
WHY NEW RELIC
Pn-depth, up-to-the-minute,
system-wide performance
data in a single solution
CASE STUDY: CANVASPOP
Challenges
Like any online retail business, CanvasPop experiences major spikes
in traffic — some more predictable than others. Most of the site’s traffic
appears during the fourth-quarter holiday season, when average
pageviews rise from about 25,000 per day to as many as 70,000 per
day. At other times of the year, heavy media coverage may cause
similar jumps in demand. “When we were on Good Morning America,
our traffic jumped to 10 times the daily average,” says Paul Brohman,
Lead Software Developer at CanvasPop. “We love the exposure.
But the exposure won’t do us any good if our site can’t handle those
unexpected visitors.”
Another challenge arises in the CanvasPop manufacturing facility,
where order fulfillment requires a complex interchange between
digital and physical processes. When a customer creates an order,
he or she will upload a digital image and select from a variety of
canvas sizes, frame designs, and image effects. Then the CanvasPop
system delivers the image to an in-house graphic artist who can
work one-on-one with the customer to recommend adjustments or
enhancements. After printing on canvas, the customer’s digital image
becomes a physical product, with CanvasPop employees laminating
and then manually framing each piece of artwork prior to shipping.
“The interplay between digital image and physical product can make
order fulfillment very tricky,” says Patrick Leckey, Senior Developer
and Lead Systems Architect at CanvasPop. “If we’re not careful in
navigating those transitions, an order can simply disappear from the
system. That’s why the performance of our backend manufacturing
application is so important — without it, we essentially don’t have
a business.”
As it turns out, business is booming. But success brings its own IT
challenges, with customer demand threatening to exceed the scale of
the CanvasPop infrastructure. By late 2011, the IT team determined
that unless they performed a major site overhaul, the customer expe-
rience would begin to suffer. “Until very recently, CanvasPop was
largely an ASP application,” says Leckey. “It was a hodgepodge of
different systems. Our applications weren’t scalable. And we had very
little access to real time information about system performance,
because we only received reports from our application monitoring
vendor once a day — and even those reports were very limited in scope.”
In order to build a more scalable infrastructure and keep a closer eye
on system performance, the CanvasPop IT team decided to migrate
their entire system from ASP to PHP. “We literally redesigned our
whole system in a matter of months,” says Leckey. “It was a huge
undertaking. Fortunately for us, the day we began migrating was also
the day we deployed New Relic.”
CASE STUDY: CANVASPOP
“New Relic alerted us to the Amazon outage before
Amazon did. We were able to bring up another copy
of our database, along with some application instances
in AZs that weren’t affected, and get back online. In
fact, our infrastructure was back up and taking orders
faster than portions of Foursquare, Pinterest, or Twitter.”
Patrick Leckey
Senior Developer and Lead Systems Architect, CanvasPop
Solution
After a six-month process of building and testing, CanvasPop went live
with its new PHP-based system on August 1, 2012. The New Relic
PHP client was with them every step of the way. “Deploying New Relic
was a five-minute job at most,” says Brohman. “We downloaded the
Chef-specific recipe for the PHP agent, plugged it into Chef, and that
was it. With our old monitoring solution, we would’ve needed to wait
until the end of the day to start seeing numbers. But New Relic started
giving us actionable information within minutes.”
It wasn’t long before the CanvasPop IT team realized that they could
use New Relic to monitor not just the new PHP system, but legacy
systems as well. “It can be tough to troubleshoot problems in code
developed by someone else, so our older applications rarely got the
attention they needed,” says Brohman. “We’re decommissioning
a lot of those systems now, which gives us even less incentive to
significantly redo them. In the meantime, New Relic performs a lot
of monitoring and debugging to keep that legacy environment
performing at higher levels than ever.”
New Relic plays a crucial role in CanvasPop’s continuous deployment
strategy, providing the real time data necessary for engineers to deploy
changes to production an average of three or four times every day.
“New Relic is so effective because it identifies problems so early in the
deployment process,” says Leckey. “This is a single, unified solution
that gives us the same interface and reports whether we’re in dev,
test, or prod. And that helps us anticipate many production issues that
otherwise wouldn’t show up in testing.”
Whether monitoring the dot-com frontend or the manufacturing backend,
New Relic gives CanvasPop the tools and features necessary to keep
all systems on track. For the customer-facing eCommerce application,
the following features prove especially helpful:
• Real User Monitoring. “When we anticipate a spike in traffic, we
use RUM to see what’s going on in real time with every single process
in our system,” says Brohman. “It tells us whether or not our current
configuration can handle the load. Then we can deploy new instances
with one command line or the click of a button.”
• Transaction Traces. “Google Analytics can be helpful, but it’s really
more of a marketing dashboard,” says Leckey. “The web transactions
in New Relic are more engineering-focused, allowing us to see the
slow points in terms of database memcache or rendering.”
• App Map. “We do a lot of linking with Facebook, Instagram, PayPal,
and other external services,” says Leckey. “The App Map is invaluable
because it helps us identify if a pain point is coming from us or from
one of our third-party partners.”
Meanwhile, the manufacturing and order fulfillment application relies
on the following New Relic features:
• Slow Query Log. “With our backend app, queries tend to be more
complex than on the frontend — each query is hitting a lot more
tables,” says Leckey. “Within minutes of deploying New Relic, we
were able to spot slow queries right off the bat.”
CASE STUDY: CANVASPOP
• Error Reporting. “We obviously don’t want customers to see errors
on the frontend,” says Brohman. “But if they do, it’s usually a quick
fix. If an error happens in the backend and we lose an item in the
manufacturing process, that’s far worse. We simply can’t have errors
in our order fulfillment system, and New Relic helps us keep those
to a minimum.”
Results
With help from New Relic, CanvasPop’s application performance isn’t
just faster — it’s more consistent, too. “These days, we don’t see any
big spikes or dips in how fast our servers respond,” says Leckey. “We
have instant access to performance metrics, so we can keep pace
with sudden changes in demand. And because the alerting process
in New Relic isn’t a binary switch, we can easily see degradation of
performance before there’s a real problem. We can set the thresholds
that make sense for our system.”
When Amazon Web Services experienced a major outage on the East
Coast of the United States in October 2012, CanvasPop was able to
stay one step ahead and avoid a significant service interruption. “New
Relic alerted us to the Amazon outage before Amazon did,” says
Leckey. “We were able to bring up another copy of our database,
along with some application instances in AZs that weren’t affected,
and get back online. In fact, our infrastructure was back up and
taking orders faster than portions of Foursquare, Pinterest, or Twitter.”
In addition to providing up-to-the-minute metrics for proactive problem-
solving, New Relic also alerted the CanvasPop IT team to longstanding
issues that had simply gone unnoticed. “There’s a query that doesn’t
get hit very often in development and staging — it’s very specific to
individual customers,” says Brohman. “With New Relic, we were able
to see that as soon as the query went into production, it would slow
down into the milliseconds. By simply changing from left to inner joins,
we dropped that query from milliseconds to microseconds, which
saved us huge amounts of compute resources on our database.”
With faster, easier access to performance metrics — and with all
metrics generated by a single solution — New Relic contributes to
a major boost in productivity at CanvasPop. “By using one solution
instead of four or five different solutions, we’re shaving off hours per
week for every developer,” says Leckey. “And of course the developers
appreciate no longer needing to spend so much time hunting for
problems. With this data, they can go straight to the source of the
trouble immediately.” All of which will be critical to CanvasPop’s 2013
plans of expanding into the European market.
New Relic also contributes to knowledge-sharing across the CanvasPop
organization. “We can easily embed any chart from New Relic into
CASE STUDY: CANVASPOP
“We can easily embed any chart from New Relic into
another application. That’s a huge value-add, because
some of our colleagues may not be familiar with this
software, and they may not be interested in the nitty-
gritty technical details. But we can give them access
to key metrics by exporting relevant charts and graphs
to the applications they’re already familiar with.”
Paul Brohman
Senior Software Developer, CanvasPop
© Copyright 2013, New Relic, Inc. All rights reserved. All trademarks, trade names, service marks and logos referenced herein belong to their respective companies. 1.24.13
CASE STUDY: CANVASPOP
About New Relic
New Relic, Inc. is the premier all-in-one Saas cloud application performance management solution. We provide comprehensive and actionable
visibility into your web & mobile data centers. Our agile Saas platform combines real user monitoring, web monitoring, server monitoring, and iOS
and Android mobile monitoring in one powerful dashboard experience. Our 35,000
+
developer and operations clients use our cloud solution every
day to optimize over 58 billion daily performance metrics.
another application,” says Brohman. “That’s a huge value add, because
some of our colleagues may not be familiar with this software, and
they may not be interested in the nitty-gritty technical details. But we
can give them access to key metrics by exporting relevant charts and
graphs to the applications they’re already familiar with.”
Soon enough, that kind of knowledge-sharing will move to a whole new
level, with the installation of dashboards throughout the CanvasPop
office displaying key metrics for all employees and visitors to see. “In
many ways, application performance is the heartbeat of our business,”
says Leckey. “By using these dashboards to display, for instance, the
number of database queries we’ve run in the past 24 hours, we can
give our colleagues an up-to-the-minute understanding of the health
and vitality of the company. Broader awareness means better
alignment across the organization. And I credit New Relic with giving
us the most accurate, most current data we’ve ever had for moving
this business forward.”
Application
performance faster
and more consistent